Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. A method of estimation of a radiation exposure by biodosimetry in a sample of cells, said sample of cells prepared for cytogenetic analysis from a single individual, said method performed using an automated digitally controlled microscope system, said method comprising: (i) acquiring images of cells sequentially, wherein said images of cells contain metaphase chromosomes, and further wherein said images of cells are acquired by using said automated digitally controlled microscope system, the microscope system having a microscope with a computer-controlled digital camera, (ii) digitally analyzing objects in each image of said images of cells to determine a property or properties of segmented objects therein, said property or properties including object count, length, width, contour finite difference. and centromere density, (iii) selecting or rejecting said each image of said images based on said property or properties determined in the preceding step, thereby creating a set of selected digital images, (iv) directing the microscope system to discontinue the acquisition of images of step (i) after a sufficient number of images have been captured to determine a radiation dose, thereby generating a set of images containing metaphase chromosomes, (v) creating a set of likely dicentric chromosomes by classifying likely dicentric chromosomes in the set of selected digital images from step (iii), and determining a count of the likely dicentric chromosomes in the set of selected digital images. (vi) determining which chromosomes of the set of likely dicentric chromosomes from step (v) are not true dicentric chromosomes using segmentation procedures that discriminate true positive dicentric chromosomes from other objects, thereby identifying false positive dicentric chromosomes and determining a count of false positive dicentric chromosomes in the set of selected digital images. (vii) eliminating the set of selected digital images of false positive dicentric chromosomes from the set of likely dicentric chromosomes, (viii) determining a numerical count of false positive dicentric chromosomes and determining a count of the dicentric chromosomes in each digital image by subtracting the number of false positive dicentric chromosomes from a total number of the likely dicentric chromosomes in each image, (ix) determining a dose response for the sample, said dose response being an average dicentric chromosome frequency over all images from the sample, by summing the total number of corrected dicentric chromosomes in said set of images containing metaphase chromosomes from the sample and dividing by the number of images in said set of images containing metaphase chromosomes, (x) computing the radiation exposure using a previously determined dose response related calibration curve that is related to the dose response by the quadratic equation, Y=aX 2 +bX+c wherein a, b, and c are coefficients of the curve, and wherein X denotes dose response and Y denotes radiation exposure, (xi) sending a signal to the digitally controlled microscope system indicating that the process of collecting images from a sample has been completed, and terminating the collection of new image data for that sample.
This invention relates to a method for estimating radiation exposure in a sample of cells using biodosimetry and automated microscopy. The method addresses the challenge of accurately detecting and quantifying dicentric chromosomes, which are indicators of radiation exposure, in a high-throughput manner. The process begins by acquiring images of metaphase chromosomes from a cell sample using an automated, digitally controlled microscope system equipped with a computer-controlled digital camera. Each image is digitally analyzed to determine properties such as object count, length, width, contour finite difference, and centromere density. Based on these properties, images are selected or rejected to create a set of high-quality digital images. The microscope system stops image acquisition once a sufficient number of images are captured to determine the radiation dose. The method then classifies likely dicentric chromosomes in the selected images and counts them. To improve accuracy, it further refines this count by identifying and eliminating false positives using segmentation procedures that distinguish true dicentric chromosomes from other objects. The corrected count of dicentric chromosomes is used to calculate an average dicentric chromosome frequency across all images, which serves as the dose response. This dose response is then applied to a pre-determined calibration curve, defined by a quadratic equation (Y = aX² + bX + c), to compute the radiation exposure. Finally, the system signals the completion of image collection for the sample, terminating further data acquisition. This approach enhances the precision and efficiency of radiation dose estimation in biodosimetry.
2. The method of estimation of radiation exposure by biodosimetry of claim 1 , said classification of a predicted dicentric chromosome, c*, in a metaphase cell digital image as a false positive dicentric chromosome, where {c 1 , . . . , c N } denotes the set of N chromosomes within the image, said predicted dicentric chromosome fulfilling any one or more of the following conditions, which are performed either independently or in combination: (i) classifying a predicted dicentric chromosome, c*, as a false positive dicentric chromosome, if the pixel area, A(c), occupied by the chromosome, is related to the areas of all other chromosomes in the same metaphase cell according to: A(c*)/median({A(c 1 ), . . . , A(c N )})<0.74 (ii) classifying a predicted dicentric chromosome, c*, as a false positive dicentric chromosome, in which W mean (c) denotes the mean value of the width profile of chromosome c, and W mean (c*)/median({W mean (c 1 ), . . . , W mean (c N ) })<0.80, (iii) classifying a predicted dicentric chromosome, c*, as a false positive dicentric chromosome, in which W med (c) denotes the median value of the width profile of chromosome c, and W med (c*)/median({W mean (c 1 ), . . . , W mean (c N ) })<0.77, (iv) classifying a predicted dicentric chromosome, c*, as a false positive dicentric chromosome, in which W max (C) denotes the maximum value of the width profile of chromosome c, and W max (c*)/median({W max (c1), . . . , W max (c N ) })<0.83, (v) classifying a predicted dicentric chromosome, c*, as a false positive dicentric chromosome, in which W cent (c) (denote the width of chromosome c at the top-ranked centromere candidate, and W cent (c*)median({W cent (c 1 ), . . . , W cent (c n )})<0.72, (vi) classifying a predicted dicentric chromosome, c*, as a false positive dicentric chromosome, in which S(c) denotes the pair of side lengths of the minimum bounding rectangle enclosing the contour of chromosome c, and 1−min(S(c*))/max(S(c*))<0.28, (vii) classifying a predicted dicentric chromosome, c*, as a false positive dicentric chromosome, in which L(c) denotes the pair of arc lengths of contour halves produced by partitioning the contour of chromosome c at its centerline endpoints, and min(L(c*))/max(L(c*))<0.51, (viii) classifying a predicted dicentric chromosome, c*, as a false positive dicentric chromosome, in which L c (c) denotes the pair of arc lengths of the contour regions of chromosome c that run between the traceline endpoints of its top 2 centromere candidates, and min(L c (c*))/max(L c (c*))<0.42.
This invention relates to biodosimetry, specifically the estimation of radiation exposure by analyzing dicentric chromosomes in metaphase cell digital images. Dicentric chromosomes, which have two centromeres, are biomarkers for radiation exposure, but automated detection systems often produce false positives. The invention provides a method to classify predicted dicentric chromosomes as false positives based on morphological and geometric features. The method evaluates a predicted dicentric chromosome against multiple criteria, including area, width profiles, centromere width, bounding rectangle side lengths, and contour arc lengths. If the chromosome meets any of the specified conditions—such as having an area less than 74% of the median chromosome area or a width profile ratio below certain thresholds—it is classified as a false positive. The approach improves the accuracy of radiation exposure estimation by reducing false positives in automated biodosimetry systems. The method can be applied independently or in combination with other criteria to enhance reliability.
3. The method of estimation of radiation exposure by biodosimetry of claim 1 , said digital analysis of images of cells from the same sample, with each image containing chromosomes from a cell in metaphase, the sample comprising M images, {I 1 , . . . , I M }, where {c 1 ,. . . , c N } denote the set of N chromosomes within image I*, and SD denotes the standard deviation function, and T denotes the threshold standard deviation value that identifies outlier images, said method, after applying filters, that either individually or combination, determines whether an image shall be removed from the sample, the digital filters comprising the following steps either individually or in combination: (i) applying the Length-width ratio filter (LW) which defines the average length-width ratio of chromosomes in an image. For a given chromosome c in a given image I containing N chromosomes, L(c,I) denotes the arc length of the centerline of c, and W mean (c,I) denotes the mean value of the width profile of c. MW(I) is defined as the mean{L(c 1 ,I)/W mean (c 1 , I), . . . ,L(C N ,I)/W mean (C N ,I)} length-width ratio. I* is removed if MW(I*)>mean{MW(I1), . . . , MW(I M )}+T×SD{MW(I 1 ), . . . , MW(I M )}, (ii) applying the Centromere candidate density filter (CD) which counts occurrences of centromere candidates in images of chromosomes. For a given chromosome c in a given image I containing N chromosomes, L(c,I) denotes the arc length of the centerline of c, and N cent (c,I) denotes the number of centromere candidates of c. CD(I) is defined as the mean{N cent (c 1 ,I)/L(c 1 ,I), . . . , N cent (c N ,I)/L(c N ,I)}. I* is removed if CD(I*)>mean{CD(I 1 ), . . . , CD(I M )}+T×SD{CD(I 1 ), . . . , CD(I M )}, (iii) applying Contour finite difference filter (FD) which represents contour smoothness of chromosomes in an image. For a given chromosome c in a given image I containing N chromosomes, WP D (c,I) denotes the set of first differences of the normalized width profile of c (range normalized to interval [ 0 , 1 ]). WD(I) is defined as the mean{mean{abs{WP D (c 1 ,I)}}, . . . , mean{abs{WP D (c N ,I)}}}. I* is removed if WD(I*)<mean{WD(I 1 ), . . . , WD(I M )}−T×SD{WD(I 1 ), . . . , WD(I M )}, (iv) applying the Total object count (Obj Count) filter, which defines the number of all objects, O, including chromosomes and non-chromosomal objects detected in an image. I* is removed if O<40or O>60, (v) applying the Segmented object count (SegObjCount) filter, which defines the number of objects processed by the gradient vector flow algorithm, O GVF , in an image. I* is removed if O GVF <35 or O GVF >50, (vi) applying the Classified object ratio (ClassifiedRatio) filter, which defines the ratio of objects recognized as chromosomes, N, to the number of segmented objects, O GVF . The stringency of this filter may be configured by adjusting the threshold of the acceptable minimum ratio to be either permissive (lower) or strict (higher), so that lower. I* is removed N/O GVF <0.6 (permissive) or 0.7 (strict).
This technical summary describes a method for estimating radiation exposure through biodosimetry by analyzing digital images of chromosomes in metaphase cells. The method processes a sample of M images, each containing N chromosomes, to identify and remove outlier images that may affect the accuracy of radiation exposure estimation. The process applies a series of digital filters, either individually or in combination, to evaluate and filter the images based on specific criteria. The Length-Width Ratio (LW) filter calculates the average length-to-width ratio of chromosomes in an image. If the mean ratio of an image exceeds the sample mean plus a threshold standard deviation, the image is removed. The Centromere Candidate Density (CD) filter counts centromere candidates per chromosome length. Images with a mean density exceeding the sample mean plus a threshold standard deviation are discarded. The Contour Finite Difference (FD) filter assesses contour smoothness by analyzing width profile differences. Images with smoothness below the sample mean minus a threshold standard deviation are removed. Additional filters include the Total Object Count (Obj Count), which removes images with fewer than 40 or more than 60 detected objects, and the Segmented Object Count (SegObjCount), which removes images with fewer than 35 or more than 50 objects processed by a gradient vector flow algorithm. The Classified Object Ratio (ClassifiedRatio) filter evaluates the ratio of recognized chromosomes to segmented objects, removing images with ratios below configurable thresholds (0.6 for permissive or 0.7 for strict settings). This method ensures high-quality image data for accurate radiation exposure estimation.
5. The method of estimation of radiation exposure levels by biodosimetry of claim 2 , wherein said digital analysis of images of cells from the same sample, with each image containing chromosomes from a cell in metaphase, the sample comprising M images, {I 1 , . . . , I M }, where {c 1 , . . . , c N } denote the set of N chromosomes within image I*, and SD denotes the standard deviation function, and T denotes the threshold standard deviation value that identifies outlier images, said method, after applying filters, that either individually or combination, determines whether an image shall be removed from the sample, the digital filters comprising the following steps either individually or in combination: (i) applying the Length-width ratio filter (LW) which defines the average length-width ratio of chromosomes in an image. For a given chromosome c in a given image I containing N chromosomes, L(c,I) denotes the arc length of the centerline of c, and W mean (c,I) denotes the mean value of the width profile of c. MW(I) is defined as the mean{L(c 1 ,I)/W mean (C 1 ,I), . . . , L(C N ,I)W mean (C N ,I)} length-width ratio. I* is removed if MW(I*)>mean {MW(I1), . . . , MW(I M )}+T×SD {MW(I 1 ), . . . , MW(I M )}, (ii) applying the Centromere candidate density filter (CD) which counts occurrences of centromere candidates in images of chromosomes. For a given chromosome c in a given image I containing N chromosomes, L(c,I) denotes the arc length of the centerline of c, and N cent (c,I) denotes the number of centromere candidates of c. CD(I) is defined as the mean{N cent (c,I)/L(c 1 ,I), . . . , N cent (c N ,I)/L(c N ,I)}. I* is removed if CD(I*)>mean{CD(I 1 ), . . . , CD(I M )}+T×SD {CD(I 1 ), . . . , CD(I M )}, (iii) applying Contour finite difference filter (FD) which represents contour smoothness of chromosomes in an image. For a given chromosome c in a given image I containing N chromosomes, WP D (c,I) denotes the set of first differences of the normalized width profile of c (range normalized to interval [0,1]). WD(I) is defined as the mean{mean{abs{WP D (c 1 ,I)}} , . . . , mean{abs{WP D (c N ,I)}}}. I* is removed if WD(I*)<mean{WD(I 1 ), . . . , WD(I M )}−T×SD {WD(I 1 ), . . . , WD(I M )}, (iv) applying the Total object count (Obj Count) filter, which defines the number of all objects, O, including chromosomes and non-chromosomal objects detected in an image. I* is removed if O<40 or O>60, (v) applying the Segmented object count (SegObjCount) filter, which defines the number of objects processed by the gradient vector flow algorithm, O GVF , in an image. I* is removed if O GVF <35 or O GVF >50, (vi) applying the Classified object ratio (ClassifiedRatio) filter, which defines the ratio of objects recognized as chromosomes, N, to the number of segmented objects, O GVF . The stringency of this filter may be configured by adjusting the threshold of the acceptable minimum ratio to be either permissive (lower) or strict (higher), so that lower. I* is removed N/O GVF <0.6 (permissive) or 0.7 (strict).
7. The method of estimation of radiation exposure levels by biodosimetry of claim 2 , said method removing false positive dicentric chromosomes from images of metaphase cells of claim 2 , and selecting metaphase images by digital analysis of images of cells from the same sample, with each image containing chromosomes from a cell in metaphase, the sample comprising M images, {I 1 , . . . , I M }, where {c 1 , . . . , c N } denote the set of N chromosomes within image I*, and SD denotes the standard deviation function, and T denotes the threshold standard deviation value that identifies outlier images, said method, after applying filters, that either individually or combination, determines whether an image shall be removed from the sample, the digital filters comprising the following steps either individually or in combination: (i) applying the Length-width ratio filter (LW) which defines the average length-width ratio of chromosomes in an image. For a given chromosome c in a given image I containing N chromosomes, L(c,I) denotes the arc length of the centerline of c, and W mean (c,I) denotes the mean value of the width profile of c. MW(I) is defined as the mean {L(c 1 I)/W mean (c 1 ,I), . . . , L(c N ,I)} length-width ratio. I* is removed if MW(I*)>mean{MW(I1), . . . , MW(I M )}+T×SD{MW(I 1 ), . . . , MW(I M )}, (ii) applying the Centromere candidate density filter (CD) which counts occurrences of centromere candidates in images of chromosomes. For a given chromosome c in a given image I containing N chromosomes, L(c,I) denotes the arc length of the centerline of c, and N cent (c,I) denotes the number of centromere candidates of c. CD(I) is defined as the mean{N cent (c 1 ,I)/L(c 1 ,I), . . . , N cent (c N I)}. I* is removed if CD(I*)>mean{CD(I 1 ), . . . , CD(I M )}+T×SD {CD(I M )}, (iii) applying Contour finite difference filter (FD) which represents contour smoothness of chromosomes in an image. For a given chromosome c in a given image I containing N chromosomes, WP D (c,I) denotes the set of first differences of the normalized width profile of c (range normalized to interval [0,1]). WD(I) is defined as the mean{mean{abs{WP D (c 1 ,I)}}, . . . , mean{abs{WP D (c N ,I)}}}. I* is removed if WD(I*)<mean{WD(I 1 ), . . . , WD(I M )}−T×SD {WD(I 1 ), . . . , WD(I M )}, (iv) applying the Total object count (ObjCount) filter, which defines the number of all objects, O, including chromosomes and non-chromosomal objects detected in an image. I* is removed if O<40 or O>60, (v) applying the Segmented object count (SegObjCount) filter, which defines the number of objects processed by the gradient vector flow algorithm, O GVF , in an image. I* is removed if O GVF <35 or O GVF >50. (vi) applying the Classified object ratio (ClassifiedRatio) filter, which defines the ratio of objects recognized as chromosomes, N, to the number of segmented objects, O GVF . The stringency of this filter may be configured by adjusting the threshold of the acceptable minimum ratio to be either permissive (lower) or strict (higher), so that lower. I* is removed N/O GVF <0.6 (permissive) or 0.7 (strict).
This invention relates to biodosimetry, specifically the estimation of radiation exposure levels by analyzing metaphase cell images to identify dicentric chromosomes, which are indicators of radiation damage. The method focuses on removing false positives from these images to improve accuracy. The process involves digital analysis of cell images from a sample, where each image contains chromosomes from a cell in metaphase. The sample consists of multiple images, and the method applies various digital filters to determine which images should be excluded from the analysis. These filters include the Length-Width Ratio filter, which evaluates the average length-to-width ratio of chromosomes and removes images where this ratio exceeds a threshold based on the sample's mean and standard deviation. The Centromere Candidate Density filter counts centromere candidates in chromosomes and removes images with excessive candidates. The Contour Finite Difference filter assesses chromosome contour smoothness, removing images with unusually low smoothness. The Total Object Count filter ensures the image contains between 40 and 60 objects, including chromosomes and non-chromosomal objects. The Segmented Object Count filter checks that the number of objects processed by a gradient vector flow algorithm falls between 35 and 50. Finally, the Classified Object Ratio filter verifies that a sufficient proportion of segmented objects are recognized as chromosomes, with configurable thresholds for permissive or strict filtering. By applying these filters individually or in combination, the method improves the reliability of radiation exposure estimation by reducing false positives in dicentric chromosome detection.
9. The method of estimation of radiation exposure levels by biodosimetry in a sample from an individual of claim 2 , said method wherein false positive dicentric chromosomes from images of metaphase cells of claim 2 are removed at said eliminating the set of selected digital images of false positive dicentric chromosomes step of claim 1 .
This invention relates to biodosimetry, specifically the estimation of radiation exposure levels by analyzing dicentric chromosomes in metaphase cells from a biological sample. The method addresses the challenge of false positives in radiation exposure detection, which can lead to inaccurate assessments of radiation dose. The process involves capturing digital images of metaphase cells from a biological sample, such as blood, to identify dicentric chromosomes—a hallmark of radiation exposure. However, non-radiation-induced dicentric chromosomes can produce false positives, complicating accurate dose estimation. The invention mitigates this issue by implementing an automated or semi-automated system to eliminate false positives from the selected digital images. The method includes preprocessing the images to enhance visibility of chromosomal structures, followed by automated detection of dicentric chromosomes. A filtering step removes false positives by analyzing chromosomal morphology, alignment, and other distinguishing features. The remaining true dicentric chromosomes are then quantified to estimate the radiation dose. This approach improves the reliability of biodosimetry by reducing errors from non-radiation-induced dicentrics, ensuring more accurate exposure assessments.
10. The method of estimation of radiation exposure levels by biodosimetry in a sample from an individual of claim 3 , said method selecting metaphase images according to claim 3 .
This invention relates to biodosimetry, a technique used to estimate radiation exposure levels in individuals by analyzing biological samples. The challenge addressed is accurately assessing radiation exposure by evaluating chromosomal damage in cells, particularly metaphase cells, which are cells in a specific stage of division where chromosomes are clearly visible. The method involves selecting metaphase images from a biological sample taken from an individual. The selection process focuses on identifying high-quality metaphase images that are suitable for detailed analysis. These images are then analyzed to detect and quantify chromosomal aberrations, such as breaks, exchanges, or other structural changes, which are indicative of radiation exposure. The frequency and type of these aberrations are used to estimate the radiation dose the individual has received. The method ensures that only the most relevant metaphase images are selected, improving the accuracy of the biodosimetry assessment. By refining the image selection process, the technique reduces variability and enhances the reliability of radiation exposure estimates. This approach is particularly useful in medical, occupational, or emergency settings where precise radiation dose assessment is critical for health monitoring and intervention.
11. The method of estimation of radiation exposure levels by biodosimetry in a sample from an individual of claim 4 , said method selecting metaphase images according to claim 4 , said method selecting metaphase images according to claim 4 , further comprising any or all of the following steps: (i) reducing the size of a confidence interval of the estimated exposure, wherein the size of the reduced confidence interval is less than the interval computed from the unselected set of metaphase images, (ii) reducing the dose estimation error to within 0.5 Gy of the corresponding physical radiation dose, (iii) demonstrating that dicentric chromosome counts among a set of selected metaphase images from the same sample are Poisson distributed thereby improving the quality of image data of said selected metaphase images.
12. The method of estimation of radiation exposure levels by biodosimetry of claim 1 , wherein the automatic selection of digital images obtained from metaphase cells from a sample isolated from an individual is performed by ranking images with a score computed from the known lengths of chromosomes, which are proportionate to the known base-pair counts of each complete chromosome, whereby the quality of a metaphase cell image is determined by comparing distribution of observed chromosome object lengths with the expected distribution of lengths obtained from relative known base-pair counts of chromosome in the reference human genome sequence, as follows: (i) the individual chromosome lengths in each image are approximated according to their corresponding chromosome areas in pixels, (ii) a fractional area of each chromosome relative to the total area of all chromosomes is determined, (iii) the chromosomes are binned according to base-pair lengths into categories corresponding to grouping defined by the International System of Cytogenetic Nomenclature, namely (1) groups A and B, which contain >2.9% of the DNA, (2) group C, which contains between 2 and 2.9% of DNA, and (3) groups D, E, F, and G, which contain <2% of the DNA (4) X chromosome, which contains approximately 2.9% of the DNA, and (5) Y chromosome which contains approximately 2% of the DNA, of the total base-pairs in a complete chromosome set, (iv) the thresholds in (iii) are compared to the fractional area of each chromosome in the metaphase image, accounting for the correct length of the sex chromosomes by reference to the known sex of the individual from whom the sample was obtained, by categorizing the result for each of the three bins in an image as a 3-element vector, and calculating a Euclidean distance from the vector to an idealized vector based on the reference human chromosome lengths, (v) sorting and ranking these Euclidean distances for all images in a sample, (vi) and eliminating images from a sample with the largest Euclidean distances, which exhibit the lowest similarity to the chromosome length distributions in a normal karyotype.
13. The method of estimation of radiation exposure levels by biodosimetry in a sample of an individual of claim 3 , said method further comprising at least one of steps (i)-(iii) below (i) reducing the size of a confidence interval of the estimated exposure wherein the size of the reduced confidence interval is less than the interval computed from the unselected set of metaphase images, (ii) reducing the dose estimation error to within 0.5 Gy of the corresponding physical radiation dose, (iii) demonstrating that dicentric chromosome counts among a set of selected metaphase images from the same sample are Poisson distributed, thereby improving the quality of image data of said selected metaphase images of claim 3 .
14. The method of estimation of radiation exposure levels by biodosimetry in a sample of an individual of claim 2 , said method further comprising at least one of steps (i)-(iii) below, (i) reducing the size of a confidence interval of the estimated exposure, wherein the size of the reduced confidence interval is less than the interval computed from the unselected set of metaphase images, (ii) reducing the dose estimation error to within 0.5 Gy of the corresponding physical radiation dose, (iii) demonstrating that dicentric chromosome counts among a set of selected metaphase images from the same sample are Poisson distributed thereby improving the quality of image data of said selected metaphase images.
15. The method of estimation of radiation exposure levels by biodosimetry of claim 12 , further comprising: determined by: (i) determining an observed distribution of dicentric chromosomes in all of the cell images in the sample according to the number of cells containing i dicentric chromosomes, where i=0 or an integer >0, (ii) estimating an expected distribution of dicentric chromosomes from a Poisson distribution, with the λ parameter of the distribution set to the average number of dicentric chromosomes per cell in all of the cell images in the sample, (iii) computing a Pearson Chi-squared goodness of fit statistic based on the observed and expected dicentric chromosome distributions for i-1 degrees of freedom and α=0.01, (iv) performing steps (i), (ii), and (iii) for the set of images in the sample after removal of the low quality images, (v) determining if the sample null hypothesis that the dicentric chromosomes in the sample follow a Poisson distribution is rejected for the complete set of images and accepted for the sample wherein low quality images have been removed.
16. The method of estimation of radiation exposure levels by biodosimetry of claim 12 , said method further comprising: (i) selection of a set of samples of known radiation doses, each consisting of a set of metaphase cell images, (iii) assignment of a support vector machine sigma value for dicentric chromosome detection, (iv) assignment of a maximum number of images to be ranked, (iv) assignment of a range of parameter values spanning the search space of all possible image selection models that are evaluated and compared to determine the accuracy of each combination of parameters, (v) evaluation of one or more parameter combinations either by selecting the model with the highest p-values of Poisson fit of dicentric chromosome distribution (p>0.05) for all samples in the set, or by selecting an optimal dose calibration curve from the sample set in (i) by minimizing the residual deviations from the known radiation dose, or by performing a leave-one cross-validation of the estimated dose for each of the samples in (i), (vi) presents an optimal automated selection models found during the search sorted according to the overall accuracy of dose estimation determined from the root mean squared sum of differences between the estimated and physical radiation doses over all samples in the set.
Unknown
February 23, 2021
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.